Efficient Resource Allocation Frameworks for Data and Service Delivery to Connected Vehicles in Vehicular Edge Computing

Abstract

As the world progresses rapidly in terms of innovation and technologies, one of the fields that has seen a significant change in the past few years is the Internet of Things. IoT has changed the scope of connectivity across the world. As more and more devices connect to the internet, the ease of use and level of automation increase for those devices, leading them to use the term ’smart’ device. Similarly, one sub-field of IoT which has been getting a lot of attention is the Internet of Vehicles (IoV), primarily with the idea of bringing self-driven cars to reality. However, IoV, as with IoT, has its own set of challenges, constraints, and use cases to consider. Due to the increase in connectivity and sophisticated software, modern vehicles are able to leverage different kinds of services provided by the environment. These services include but are not limited to, data delivery and computation offloading. While cloud computing was initially considered to deliver these services, it was realized that communication with the cloud would require constant high bandwidth requirements and could incur high latency issues, especially while travelling through low network regions. Thus, to solve these issues, instead of Cloud Computing, Edge Computing was considered, as it brought the required computation units closer to the user vehicles, thus requiring lesser bandwidth and facing low latency. This technology of using Edge Computing to deliver services to connected vehicles while considering their dynamic network topology is called Vehicular Edge Computing (VEC). VEC requires the use of multiple edge nodes throughout the road network to facilitate the services for connected vehicles. These edge nodes can also be connected to a central cloud server. However, VEC also has its own set of challenges and constraints to consider when delivering each type of service. Data delivery to connected vehicles would require the allocation of memory resources at the edge nodes that the vehicle will pass through and receive and store the data from the cloud to be delivered to the vehicle. It should also have enough bandwidth to transfer the data to the vehicle when it passes through the coverage region of that edge node. Computation offloading would require proper scheduling of tasks across all the edge nodes in the network so as to facilitate efficient execution of the tasks and ensure the results are delivered within the deadlines given the resource constraints. All these services have to consider multiple vehicles at the same time at the coverage of an edge. Thus, resources should be properly allocated to service the maximum number of vehicles. This work tries to address these challenges by contributing solutions considering the vehicle flow constraints of the network as well as resource and timing constraints when delivering data to the vehicles or scheduling the tasks of the vehicles for computation offloading. The first solution proposes a twostage optimization framework for efficient data delivery to connected vehicles via edge nodes while considering dynamic route changes. This framework optimizes the bandwidth utilized to send data from the cloud to the edge nodes. It also prioritizes vehicles with more data to receive and fewer edge nodes to pass through to reach their destination. The second solution proposes a Global Earliest Deadline First (GEDF) based scheduler for scheduling offloading tasks to the edge nodes. This approach considers the vehicle flow constraints, resource constraints, and timing constraints while assigning tasks to edge nodes that can execute the offloaded task, while prioritizing vehicles with shorter deadlines. Both these works have been tested on a real-world dataset (Luxembourg dataset) and compared with other approaches, including an optimal approach with various parameters such as the number of vehicles fully serviced, run time, etc